Anthropic releases Claude Opus 4.7, leading AI benchmarks in coding and agentic workflows

Anthropic has released Claude Opus 4.7, its most capable generally available AI model to date, excelling in coding, agentic workflows, and complex reasoning tasks. The update, announced on April 16, 2026, positions Opus 4.7 as a benchmark leader, narrowly reclaiming the top spot among publicly accessible large language models, according to reports from Bloomberg Technology and VentureBeat.

This new version builds on predecessors like Opus 4.6, delivering stronger performance in software engineering, multi-step instruction-following, tool use, and high-resolution image processing—up to three times higher than before. As detailed by The Next Web, Opus 4.7 scores 64.3% on the demanding SWE-bench Pro evaluation, surpassing competitors like GPT-5.4's 57.7%, while also advancing multi-agent coordination for workflows lasting hours. Anthropic's own documentation highlights its hybrid reasoning capabilities, including a 1 million token context window, enabling thorough handling of professional knowledge work with minimal oversight.

Unlike the more powerful but restricted Claude Mythos Preview—rolled out just a week earlier to select companies under Project Glasswing, a cybersecurity initiative—Opus 4.7 is designed to be less broadly capable and lower risk for wider use, as noted in Slashdot coverage citing CNBC. Mythos remains limited to enterprise partners, allowing Anthropic to test cutting-edge features safely before broader deployment. Opus 4.7, by contrast, is immediately accessible via Anthropic's API, Amazon Bedrock, Google Cloud's Vertex AI, and platforms like GitHub Copilot for enterprise users.

The release underscores intensifying competition in AI development, where rapid iterations drive advancements in autonomous agents capable of planning, executing, and learning across extended sessions. Developers and businesses stand to benefit most, with applications in coding marathons, research, scientific discovery, and production workflows—such as autonomously managing issues across repositories or orchestrating multi-tool tasks. Pricing aligns with prior Opus models at $15 per million input tokens and $75 for output, making it viable for Pro, Max, Team, and Enterprise plans.

Looking ahead, Opus 4.7's integration into major cloud services and tools like Claude Code signals a shift toward AI as reliable collaborators rather than mere assistants. Early adopters report consistent reliability on ambitious projects, from agent teams tackling complex coding to adaptive thinking that balances speed and depth. As demand for AI chips surges—evident in TSMC's recent profit jump and raised 2026 outlook—this model could accelerate enterprise adoption, though Anthropic's cautious approach with restricted previews like Mythos reflects ongoing concerns over safety and capability boundaries.